Silicon Ingot Tail Growth Control with Adaptive Neural Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The CZ method for manufacturing silicon wafers faces challenges in controlling the tail portion of the ingot due to manual setting of target values, which leads to inconsistent growth and increased process time, especially when thermal environments change.
Innovation Solution
An ingot growing apparatus and method utilizing an artificial neural network to learn and update shoulder shape models, setting target tail temperatures based on these models to control the growth of the tail portion, ensuring consistent shape and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual target value setting is used for tail portion growth, then operator experience and judgment are utilized, but manufacturing precision and consistency deteriorate due to inability to maintain slip dislocation and long reflection cycles
Solution Approach 1:
The system implements real-time feedback by continuously monitoring actual tail portion growth parameters (temperature, diameter, shape) and comparing them with target values. The control unit adjusts operating parameters dynamically based on deviations, enabling automatic maintenance of slip dislocation and consistent tail portion formation without manual intervention.
Solution Approach 2:
The control system performs self-adjustment by automatically detecting growth conditions and modifying process parameters to maintain optimal tail portion growth. The system monitors its own performance and corrects deviations autonomously, eliminating the need for operator judgment and achieving consistent results across different ingots.
2Device complexity
If manual target value setting is used, then setup simplicity is maintained, but productivity deteriorates due to long reflection cycles and increased process time
Solution Approach 1:
The system performs preliminary calculations and preparations by pre-establishing target value ranges and adjustment protocols before actual growth occurs. The control unit anticipates needed adjustments and prepares correction sequences, enabling rapid response to deviations and reducing overall process time without requiring complex real-time manual control.
Solution Approach 2:
Real-time feedback mechanisms enable the system to detect and correct deviations immediately, eliminating the long reflection cycles inherent in manual adjustment. The control unit continuously monitors growth parameters and implements automatic corrections, significantly reducing tail process time while maintaining consistent quality.
3Ease of operation
If same target value is used for tail portion growth, then operational simplicity is maintained, but manufacturing precision deteriorates because thermal environment changes cause inconsistent growth
Solution Approach 1:
The system transitions from static target values to dynamic, adaptive target values that automatically adjust according to changing thermal environments and growth conditions. The control unit modifies temperature, diameter, and shape parameters in real-time to maintain consistent tail portion geometry despite environmental variations.
Solution Approach 2:
The system dynamically changes process parameters (temperature, pulling speed, rotation speed) based on monitored growth conditions and thermal environment variations. By continuously adjusting these parameters, the system maintains optimal growth conditions and produces consistent tail portion shapes across different ingots.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The apparatus enables standardized and optimal growth of the tail portion of silicon ingots, even with changing thermal environments, reducing process time and improving productivity and yield.
Implementation Method 1
a processor configured to learn an artificial neural network to generate a primary shoulder shape model corresponding to training data, to update the primary shoulder shape model to generate a secondary shoulder shape model, and to set a target tail temperature for growing a tail portion of the ingot based on the secondary shoulder shape model
Implementation Method 2
polycrystalline silicon is charged into a quartz crucible. This is heated and melted by a graphite heating element
Implementation Method 3
a seed crystal is immersed in silicon melt formed as a result of melting, and crystallization occurs at the interface
Implementation Method 4
the tail process is a process of reducing the diameter while maintaining a slip dislocation state in the body portion of the ingot
Data Source
AI summary
An ingot growing apparatus is composed of a neck portion, a shoulder portion, a body portion, and a tail portion. The ingot growing apparatus comprises a memory configured to store an artificial neural network and a processor.The processor learns the artificial neural network to obtain a primary shoulder shape model corresponding to training data, updates the obtained primary shoulder shape model based on shoulder information obtained during the growth of a shoulder portion of a first ingot to obtain a secondary shoulder shape model, sets a target tail temperature for growth a tail portion of the first ingot based on the secondary shoulder shape model, and controls the growth of the tail portion of the first ingot according to the set target tail temperature.